Fix Series.item to run a Spark job once (#1518)

According to the comment #1502 (comment), fixed `Series.item` to run a Spark job not twice, once.

databricks/koalas/series.py

@@ -4873,10 +4873,7 @@ def item(self):
         >>> kser.item()
         10
         """
-        item_top_two = self[:2]
-        if len(item_top_two) != 1:
-            raise ValueError("can only convert an array of size 1 to a Python scalar")
-        return item_top_two[0]
+        return self.head(2).to_pandas().item()
 
     def _cum(self, func, skipna, part_cols=()):
         # This is used to cummin, cummax, cumsum, etc.